Internship project · Computer Vision Detection
The problem
Active construction environments need continuous safety oversight, but human monitoring can’t catch every violation in real time.
Approach
Developed a detection workflow that flags critical safety violations from on-site footage:
- Collects and prepares multi-dataset training data
- Trains a detection model (YOLO11m)
- Deploys detection for on-site safety checks
Stack
- YOLO11m
- Multi-dataset detection pipeline
Outcome
Monitored safety harness, smoking, crane, and fence violations — supporting on-site safety checks with automated, real-time detection.
[ screenshots + repo link coming soon ]